Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Initial classification of terrain based on 9 topographic parameters extracted from a 10 meter DEM. Slope, profile curvature, maximum curvature, minimum curvature, longitudinal curvature, and cross-sectional curvature were computed from a 15x15 moving window by r.param.scale. A 9x9 moving window was used to compute the hypsometric integral as defined by Wood and Snell (1960). These 7 parameters were extracted from a given region, and classified with the clara() function from the cluster package in R. Transferal of raster data between GRASS and R was made possible by the spgrass6 package. An excellent overview of some of the features in the spgrass6 package is included in the third volume of the GRASS newsletter. Some initial results are displayed in Figures 1 and 2. Terrain data is better visualized in 3 dimensions: Figure 3 contains a view of both Temblor and Volcanic lanforms near McCabe Canyon. Ideas for this type of classification technique came from Roger Bivand (Bivand, 2000).

Background

Four regions were selected within the PINN boundary for analysis of intrinsic slope classes, based on terrain type.Figure 1 illustrates the four regions that were choosen. Figure 2 shows histograms of each region along with intrinsic slope class breaks (red) and NRCS slope class breaks (gray). Note that NRCS slope classes were derived from a similar analysis (k-means clustering) of slope values from the entire park: e.g. without regard for specific landform. See attached PDF version of Figure 2 at the bottom of this page.

Depicting the relative density of sampling can be an interesting task, especially when the points are highly clustered and sparse. Three simple operations that can reveal subtle patterns in the spatial distribution of sample points were explored:

• A voronoi tessellation vector operation performed on the actual sample locations.
• A pseudo-density raster operation based on a count of cells, within proximity to sample points, and within a given radius.
• A kernel-smoothed density function, based on a given standard deviation of spatial density.
• A density estimation performed by the density() function from the Spatstat package in R

Voronoi tessellation

Psudo-density calculation

Kernel-smoothed density

Spatial density estimation in R

More often than not geographic data come in a format that is essentially two-dimensional: i.e. raster grids and vector points, lines, and areas with only (x,y) style coordinates.

GRASS Vectors

SSURGO cutaway

Slope class cutaway

Topomap

Panchromatic

Near infrared

Geomorphic features.

View of Mt. Defiance from Bear Valley, Pinnacles National Monument. Pan Chromatic 2m res.

Thanks to Markus Neteler for initial insight on how to build an appropriate POVRAY script file.